CN101068108A - Orthogonal mirror image filter group realizing method and device based on genetic algorithm - Google Patents
Orthogonal mirror image filter group realizing method and device based on genetic algorithm Download PDFInfo
- Publication number
- CN101068108A CN101068108A CN 200710119030 CN200710119030A CN101068108A CN 101068108 A CN101068108 A CN 101068108A CN 200710119030 CN200710119030 CN 200710119030 CN 200710119030 A CN200710119030 A CN 200710119030A CN 101068108 A CN101068108 A CN 101068108A
- Authority
- CN
- China
- Prior art keywords
- gene
- fitness
- probability
- character string
- highest
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 41
- 230000002068 genetic effect Effects 0.000 title claims abstract description 40
- 238000000034 method Methods 0.000 title claims abstract description 38
- 108090000623 proteins and genes Proteins 0.000 claims abstract description 206
- 238000005457 optimization Methods 0.000 claims description 23
- 238000012545 processing Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 12
- 230000003044 adaptive effect Effects 0.000 abstract 3
- 229910002056 binary alloy Inorganic materials 0.000 abstract 1
- 230000006870 function Effects 0.000 description 12
- UELITFHSCLAHKR-UHFFFAOYSA-N acibenzolar-S-methyl Chemical compound CSC(=O)C1=CC=CC2=C1SN=N2 UELITFHSCLAHKR-UHFFFAOYSA-N 0.000 description 11
- 210000004027 cell Anatomy 0.000 description 11
- 238000010586 diagram Methods 0.000 description 8
- 210000000349 chromosome Anatomy 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000015572 biosynthetic process Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004083 survival effect Effects 0.000 description 3
- 238000003786 synthesis reaction Methods 0.000 description 3
- 101150044345 0.5 gene Proteins 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 230000013011 mating Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 101150039622 so gene Proteins 0.000 description 1
- 238000004804 winding Methods 0.000 description 1
Images
Landscapes
- Complex Calculations (AREA)
Abstract
A method for realizing orthogonal mirror image filter set based on genetic algorithm includes generating multi-set of random number to represent parameter of orthogonal mirror image filter set, coding each set of random number to binary system character string in finite length according to current gene selection probability, decoding each character string and calculating out adaptive degree, updating said probability according to gene of character string with highest adaptive degree, repeating said step till ending condition is satisfied, obtaining parameter of said filter set by decoding character string with highest adaptive degree and utilizing obtained parameter to realize said filter set.
Description
Technical field
The present invention relates to Genetic Algorithm Technology, relate in particular to a kind of implementation method and quadrature mirror filter bank Parameter Optimization device of the quadrature mirror filter bank based on genetic algorithm.
Background technology
(Genetic Algorithms GA) is heredity and darwinian evolution theory and the parallel random optimization algorithm of proposition its objective is in order to obtain optimal solution naturally in the simulation biology to genetic algorithm.In genetic algorithm, all parameters of optimization problem all are encoded, and forming one has limit for length's string of binary characters, and this character string is called " individuality ".The basic genetic unit of bion is a gene in the living nature, gene are arranged as chromosome in certain sequence, in genetic algorithm, chromosome is corresponding to the individual limited long character string of expression, and gene are corresponding to the value of each bit of character string, be generally 0 or 1, promptly represent with string of binary characters.The set of all " individualities " becomes colony.Each " individuality " is corresponding to a feasible solution of optimization problem in " colony ".The target function of optimization problem is as the residing environment of colony, and target function value is corresponding to the fitness of individuality to environment.Evolution Theory according to Darwin survival of the fittest, individuality in the colony carries out struggle for existence, the individual survival that the environmental adaptation degree is high is got off, carry out mating (being crossing operation) simultaneously and variation (being the halmatogenesis computing) raises up seed, and the low individuality of environmental adaptation degree is eliminated by colony gradually.Crossing operation is to be the crosspoint with certain bit, the portion gene on the chromosome of two individualities exchanged, like this, if after two high parents of fitness exchanged hereditary information, the offspring of generation just might have better fitness; The halmatogenesis computing is according to certain probability, and the gene of a certain position on the individual chromosome is perturbed, and makes it become oppositional gene.Gene is that 1 oppositional gene is 0, and vice versa.So evolve, last whole colony will be adapted to residing environment most, thereby obtains the optimal solution of problem generation upon generation ofly.
Referring to Fig. 1, in the prior art genetic algorithm being applied to the schematic diagram of optimization problem, GA is applied to optimization problem and generally comprises following steps:
Genetic algorithm has been used to find the solution some problems that have application prospect, for example the design of QMF (QuadratureMirror Filter, quadrature mirror filter) group, function optimization, Computer Image Processing and robot motion planning or the like.
The QMF group is one group of filter with a common input signal or a common output signal.The bank of filters that wherein has a common input signal and K output signal is called analysis filterbank; Otherwise the bank of filters with K input signal and an output signal is called the synthesis filter group.In analysis filterbank one side, input signal (being made as broadband signal) is divided into K sub-frequency bands signal (narrow band signal), can reduce sample rate by extracting; In synthesis filter one side,, can rebuild original signal by null value interpolation and bandpass filtering.
In the existing genetic algorithm, all will store and intersect and the computing that make a variation the individuality that iterative process each time produces, so exist amount of calculation big, the problem that memory space is big be unfavorable for the real-time implementation of algorithm.When therefore this genetic algorithm being applied to the QMF group, QMF group reconstruction error is strengthened, produce bigger amplitude distortion.
Summary of the invention
The invention provides a kind of quadrature mirror filter bank implementation method based on genetic algorithm, to realize reducing amplitude distortion, this method comprises the steps:
Generate many groups at random and represent the random numbers of quadrature mirror filter bank parameter, and be the string of binary characters of finite length with every group of random number code according to current gene selection probability; Each character string of decoding is also calculated fitness, upgrades described gene according to the genotype of the highest character string of the fitness that calculates and selects probability;
Repeat above-mentioned steps, up to satisfying termination condition;
The character string that fitness is the highest since successive dynasties has been decoded, obtained the parameter of described quadrature mirror filter bank, and adopt described parameter to realize quadrature mirror filter bank.
The present invention also provides a kind of quadrature mirror filter bank Parameter Optimization device, and this device comprises:
The random number generation unit is used for generating at random the random number that many groups are represented the quadrature mirror filter bank parameter;
The gene code unit is used for selecting the current gene of probability updating block to select probability according to gene, and every group of random number code that described random number generation unit is generated is the string of binary characters of finite length;
The fitness computing unit, the described string of binary characters of the described gene code unit that is used for decoding also calculates fitness, and the string of binary characters of high fitness that will calculate stores described memory cell into;
Memory cell has been used to store the highest string of binary characters of fitness since the successive dynasties;
Gene is selected the probability updating block, and the genotype that is used for the string of binary characters of the highest fitness that calculates according to described fitness computing unit is upgraded gene and selected probability, and sends decision instruction to described judgment processing unit;
The judgment processing unit is used for judging whether to satisfy termination condition according to described decision instruction, if, then decode having the string of binary characters of high fitness in the described memory cell, obtain the parameter of described quadrature mirror filter bank; Otherwise, start described random number generation unit and generate random number.
The present invention is by in the process that realizes quadrature mirror filter bank, adopt gene to select probability encoding to represent the random number of quadrature mirror filter bank parameter, thereby generation string of binary characters, and by upgrading gene selection probability, it is tilted to the highest idiotype of previous generation fitness, select character string behind the probability encoding to inherit the outstanding gene of previous generation thereby make, and obtained the parameter of quadrature mirror filter bank by the character string decoding of the highest fitness since the successive dynasties with gene.This shows, need not write down all character string genotype from generation to generation among the present invention, and as long as therefore record genotype of high fitness in per generation can reduce the calculating required storage; And, do not need among the present invention character string is intersected and the computing that makes a variation, thereby simplified implementation procedure.Therefore help the real-time implementation of parameter optimization, and reduced the amplitude distortion of quadrature mirror filter bank thus.
Description of drawings
Fig. 1 is in the prior art being applied to genetic algorithm the schematic diagram of optimization problem;
Fig. 2 is the schematic flow sheet of the genetic algorithm of the embodiment of the invention;
Fig. 3 is the 2 passage QMF group principle schematic of the embodiment of the invention;
Fig. 4 is the QMF group realization flow schematic diagram based on genetic algorithm of the embodiment of the invention;
Fig. 5 A and Fig. 5 B are the QMF group parameter optimization apparatus structure schematic diagram based on genetic algorithm of the embodiment of the invention.
Embodiment
Below in conjunction with accompanying drawing the embodiment of the invention is described in detail.
In order to make the nearly target of accurately rebuilding of QMF winding, can to reduce QMF group reconstruction error, reach the target of rebuilding by optimizing the amplitude distortion algorithm near accurately.The embodiment of the invention provides a kind of QMF group implementation method based on genetic algorithm, thereby reaches the purpose that minimizes reconstruction error.
Referring to Fig. 2, be the schematic flow sheet of the genetic algorithm of the embodiment of the invention, concrete steps comprise:
When initial, it all is P that the gene on all bits is selected the probability of " 1 "
i=0.5 (promptly get " 1 " or " 0 " is equiprobable, the gene in the present embodiment selects probable value for selecting 1 probable value, and i is an i gene in the gene strand).Gene selects the adjusted value of probability to be provided with according to population scale and iterations, and principle is to guarantee suitable convergence rate, and the gene in the present embodiment selects the adjusted value of probability to be set to ε.
Present embodiment is controlled iterations according to generation number,, evolutionary generation as the Rule of judgment that whether finishes to calculate, therefore also need be provided with the initial value and the threshold value of generation number in initialization procedure that is.The initial value of generation number can be set to 0, and the threshold value of generation number can be according to population scale, and the computational accuracy that needs is provided with.
For example, generate N group random number according to population scale, every group of random number number is M, and the scope of random number is 0～1.Wherein, one group of random number is represented body one by one, and N group random number represents to have generated N individuality; M represents each individual gene number.
For example, certain individual random number sequence that generates is R={r[0], r[1] ..., r[M-1] }, the binary sequence behind the coding be R '=r ' [0], r ' [1] ..., r ' [M-1], then select probability P to be for this individual coding method according to gene:
If r[i]＜P[i], then: r ' [i]=1;
Otherwise: r ' [i]=0, i=0～M-1
Because r[i] be the random number between 0～1, so gene selects probability P [i] more near 1, then this random number might be encoded as 1 more.
If working as the former generation individuality is first generation individuality, then the selection of the gene on all bits probability all is P
i=0.5, like this, when using gene to select probability to when the gene of former generation individuality is encoded, then each gene position is encoded as 1 or 0 will be equiprobable to gene selection probability.
If when the former generation individuality is the later individuality of the second generation, then gene selection probability might be different with initial value, the selection probability that is some gene position may be greater than 0.5, and the selection probability of some gene position may be less than 0.5, like this, select probability to when the gene of former generation individuality is encoded when using gene, select probability greater than 0.5 gene position for gene, then this gene position is encoded as 1 probability will increase, and the gene of this gene position selects probability big more, and then to be encoded as 1 possibility also big more for this gene position.In like manner, select probability less than 0.5 gene position for gene, then this gene position is encoded as 0 probability will increase, and the gene of this gene position selects probability more little, and then to be encoded as 0 possibility big more for this gene position.
Through such cataloged procedure, can make the individuality that generates at random in the above-mentioned steps more approach to select the excellent individual of probability decision by gene.
Corresponding relation according to genetic algorithm and optimization problem is decoded, and calculates each individual fitness.
The genotype of step 206, the basis individuality that fitness is the highest in the former generation individuality is adjusted and is upgraded gene selection probability, and generation number is added one, returns step 202 then, proceeds iterative computation.
The highest individuality of fitness is the optimum individual when former generation in the former generation individuality, preserves this optimum individual.According to the selection probability that upgrades gene when the genotype of former generation optimum individual, promptly the genotype with optimum individual is reference, and the gene of all bits selects probability to tilt to it.
For example, the genotype of i gene position of optimum individual is ' 1 ', then allows this gene position select probability to be partial to it, promptly revises P
i, make that the probability of this gene position selection ' 1 ' is P
i=P
i+ ε, wherein ε is the adjusted value that gene is selected probability, is an a small amount of, when generating offspring individual, the probability of this gene position choosing ' 1 ' will increase like this; Equally, if certain gene position of optimum individual is ' 0 ', then allowing this gene position select ' 1 ' probability is P
i=P
i-ε, according to the method described above, with the selection probability P of all gene position
iAll upgrade.
From above-mentioned flow process as can be seen, select probability by upgrading gene, and the genes of individuals position of selecting probability encoding to generate at random with the gene after upgrading, outstanding genotype is better inherited, bad genotype is eliminated gradually, like this, and through after some generations, the selection probability of each gene position will converge to 1 or 0, and this moment, iteration work just can finish.Each is for the just optimal solution of correspondence problem of global optimum's individuality of optimum individual.
Do not store each in the above-mentioned flow process for all individual gene chain codes, but only preserve the highest genes of individuals chain code of fitness in the contemporary individuality, thereby reduced memory space.Owing to do not write down each for all individual genes, thereby intersect and make a variation and to realize.But the purpose of considering intersection is to make the offspring inherit parents' good characteristic, thereby biocenose is developed towards the high direction of fitness; The purpose of variation is to avoid biological " precocity ", promptly sinks into the suboptimum hereditary capacity too early.In above-mentioned flow process, the effect that intersects and make a variation selects probability to replace finishing by gene, because before producing bion of new generation, gene selects probability all can tilt to the optimum bion genotype of previous generation, thereby helps the good characteristic that bion of new generation is inherited previous generation to a certain extent.Make biocenose develop towards the high direction of fitness.Simultaneously, gene selects probability to tilt to optimum bion, wherein optimum bion is the optimum bion of previous generation, rather than the optimum bion since the successive dynasties, therefore, though the fitness of the optimum bion of previous generation may be lower than the fitness of the optimum bion since the successive dynasties, but reduced to have selected the inclination of probability with reference to the problem that converges to locally optimal solution easily too early that is caused for gene as each with the optimum bion since the successive dynasties as far as possible.This shows that gene selects the effect of probability can substitute intersection and variation fully, and can reduce computational complexity and storage capacity requirement greatly.
Present embodiment is an example with 2 passage QMF group, describes and adopts above-mentioned improved genetic algorithm to realize the process of QFM group.In the schematic diagram of 2 passage QFM group as shown in Figure 3, H0 and H1 composition analysis bank of filters, F0 and F1 form the synthesis filter group: according to general QMF group method for designing, can draw the relation of each filter function of 2QMF:
The time domain coefficient of corresponding 2QMF filter is:
Improved genetic algorithm is applied to the design of QMF group, promptly on the basis of general QMF group method for designing, further utilizes above-mentioned improvement genetic algorithm to carry out parameter optimization, to reach the purpose that minimizes reconstruction error.
Referring to Fig. 4, be in the embodiment of the invention based on the realization flow schematic diagram of the QMF of improved genetic algorithm group, concrete steps comprise:
The target function (reflection error) of step 401, definition genetic optimization problem, wherein,
Reconstruction error is:
The passband error is:
The stopband error is:
H0 wherein, the frequency characteristic of H1 is normalization all, ω
dBe the filter excess bandwidth.
Target function is: Φ=α Φ
1+ β Φ
2+ γ Φ
3, wherein alpha+beta+γ=1 is weighted factor.
The parameter to be optimized that step 402, definition utilize above-mentioned improved genetic algorithm to be optimized.
In the present embodiment, set the to be optimized parameter of QMF group filter h0 coefficient for utilizing improved genetic algorithm to be optimized.
The initialization procedure of step 403, genetic algorithm.
The initialization procedure of genetic algorithm is with the step 201 among Fig. 2, and initialized parameter comprises in the present embodiment: the initialization gene is selected probability P [i]=0.5, and i=0～M-1, M are each genes of individuals number.
For example, QMF group filter h0 coefficient length is L, and each coefficient quantizes with 8bit, and the bit number that then needs altogether is M=8L, and promptly in the gene string of h0 individuality, per in order 8 genes can decode the coefficient of h0.
If as the Rule of judgment that calculates termination, then in initialization procedure, need to be provided with generation number initial value and threshold value with generation number; If whether reach certain threshold value as the Rule of judgment that calculates termination with target function value, then in initialization procedure, need to be provided with the target function value threshold value.
Select probability according to random number that produces and gene, for this generation biological gene is encoded, the coding back produces the string of binary characters that length is M, and coding method is with flow process shown in Figure 2.
According to the parameter value scope, each individuality is decoded, obtain each individual corresponding h0, through type 3.2 is tried to achieve h1, f0, f1, and then can calculate each individual fitness according to the target function that provides in the step 401, promptly reflect the target function value of error.
According to the gene type of the highest biology of contemporary fitness, upgrade gene and select probability P [i], gene is selected the description of the update method of probability with Fig. 2 flow process.If, then also will add one to generation number with the Rule of judgment of generation number as the calculating termination.Return step 404 then.
Because the genetic algorithm that adopts in the above-mentioned flow process is a kind of algorithm that solves optimization problem, use at the design of QMF group, obtain the optimal solution of problem by the most optimized parameter.It is filter coefficient that genetic algorithm needs optimum parameters, and the fitness of genetic algorithm is relevant with the amplitude distortion of bank of filters.Be that fitness is high more, amplitude distortion is more little.Therefore, when adopting this genetic algorithm to obtain the genotype of the highest fitness, promptly obtained making the filter coefficient of QMF group amplitude distortion minimum.
The embodiment of the invention also provides two kinds of QMF group Parameter Optimization devices.
Referring to Fig. 5 A, be the QMF group parameter optimization apparatus structure schematic diagram based on genetic algorithm of the embodiment of the invention, this device comprises: random number generation unit, gene code unit, fitness computing unit, memory cell, gene are selected probability updating block, judgment processing unit and counting unit.
The random number generation unit is used for generating the random number that many groups are represented the quadrature mirror filter bank parameter at random, and the notifications count unit adds one with count value;
The gene code unit is used for selecting the current gene of probability updating block to select probability according to gene, and every group of random number code that the random number generation unit is generated is the string of binary characters of finite length;
The fitness computing unit be used for decoding the gene code unit string of binary characters and calculate fitness, and the string of binary characters of high fitness that will calculate stores memory cell into;
Memory cell has been used to store the highest string of binary characters of fitness since the successive dynasties;
The genotype that gene selects the probability updating block to be used for the string of binary characters of the highest fitness that calculates according to the fitness computing unit is upgraded gene and is selected probability, and sends decision instruction to the judgment processing unit.It is 0.5 that gene selects the initial gene of each gene position in the probability updating block to select probability.When gene selects the probability updating block to upgrade gene selection probability, select the genic value of probability corresponding gene position in the highest character string of fitness that the fitness computing unit calculates to tilt the gene of each gene position, for example, genic value as if corresponding gene position in the highest character string of current fitness is 1, and then gene selects the probability updating block to select probability to increase the numerical value of appointment the gene of corresponding gene position; Genic value as if corresponding gene position in the highest character string of current fitness is 0, and then gene selects the probability updating block to select probability to reduce the numerical value of appointment the gene of corresponding gene position; Perhaps, if the genic value of corresponding gene position is 1 in the highest character string of current fitness, then gene selects the probability updating block to select probability to reduce the numerical value of appointment the gene of corresponding gene position; Genic value as if corresponding gene position in the highest character string of current fitness is 0, and then gene selects the probability updating block to select probability to increase the numerical value of appointment the gene of corresponding gene position.
The judgment processing unit is used for selecting according to gene the decision instruction of probability updating block transmission, read the count value in the counting unit, if count value is more than or equal to the threshold value that sets in advance, then judge the current termination condition that satisfied, then decode, obtain the parameter of quadrature mirror filter bank having the string of binary characters of high fitness in the memory cell; If count value, then starts the random number that the random number generation unit generates a new generation less than the threshold value that sets in advance.
Referring to Fig. 5 B, the another kind that provides for the embodiment of the invention is based on the QMF group parameter optimization apparatus structure schematic diagram of genetic algorithm, and this device comprises: random number generation unit, gene code unit, fitness computing unit, memory cell, gene are selected probability updating block, judgment processing unit and fitness comparing unit.
The random number generation unit is used for generating at random the random number that many groups are represented the quadrature mirror filter bank parameter;
The gene code unit is used for selecting the current gene of probability updating block to select probability according to gene, and every group of random number code that the random number generation unit is generated is the string of binary characters of finite length;
The fitness computing unit be used for decoding the gene code unit string of binary characters and calculate fitness, and the string of binary characters of high fitness that will calculate stores memory cell into;
Memory cell has been used to store the highest string of binary characters of fitness since the successive dynasties;
The fitness that the fitness comparing unit is used for the highest string of binary characters of fitness that the fitness computing unit is calculated compares with default fitness threshold value;
The genotype that gene selects the probability updating block to be used for the string of binary characters of the highest fitness that calculates according to the fitness computing unit is upgraded gene and is selected probability, and sends decision instruction to the judgment processing unit.It is 0.5 that gene selects the initial gene of each gene position in the probability updating block to select probability.When gene selects the probability updating block to upgrade gene selection probability, select the genic value of probability corresponding gene position in the highest character string of fitness that the fitness computing unit calculates to tilt the gene of each gene position, for example, genic value as if corresponding gene position in the highest character string of current fitness is 1, and then gene selects the probability updating block to select probability to increase the numerical value of appointment the gene of corresponding gene position; Genic value as if corresponding gene position in the highest character string of current fitness is 0, and then gene selects the probability updating block to select probability to reduce the numerical value of appointment the gene of corresponding gene position; Perhaps, if the genic value of corresponding gene position is 1 in the highest character string of current fitness, then gene selects the probability updating block to select probability to reduce the numerical value of appointment the gene of corresponding gene position; Genic value as if corresponding gene position in the highest character string of current fitness is 0, and then gene selects the probability updating block to select probability to increase the numerical value of appointment the gene of corresponding gene position.
The judgment processing unit is used for selecting according to gene the decision instruction of probability updating block transmission, read the comparative result of fitness comparing unit, if the highest fitness is equal to or greater than the fitness threshold value that sets in advance, then judge the current termination condition that satisfied, then decode, obtain the parameter of quadrature mirror filter bank having the string of binary characters of high fitness in the memory cell; If the highest fitness less than default fitness threshold value, then starts the random number that the random number generation unit generates a new generation.
Obviously, those skilled in the art can carry out various changes and modification to the present invention and not break away from the spirit and scope of the present invention.Like this, if of the present invention these are revised and modification belongs within the scope of claim of the present invention and equivalent technologies thereof, then the present invention also is intended to comprise these changes and modification interior.
Claims (11)
1, a kind of implementation method of the quadrature mirror filter bank based on genetic algorithm is characterized in that, comprises the steps:
Generate many groups at random and represent the random numbers of quadrature mirror filter bank parameter, and be the string of binary characters of finite length with every group of random number code according to current gene selection probability; Each character string of decoding is also calculated fitness, upgrades described gene according to the genotype of the highest character string of the fitness that calculates and selects probability;
Repeat above-mentioned steps, up to satisfying termination condition;
The character string that fitness is the highest since successive dynasties has been decoded, obtained the parameter of described quadrature mirror filter bank, and adopt described parameter to realize quadrature mirror filter bank.
2, the method for claim 1 is characterized in that, it is 0.5 that the initial gene of each gene position is selected probability.
3, the method for claim 1 is characterized in that, upgrades the process that described gene is selected probability according to the genotype of the highest character string of the current fitness that calculates, and comprises step:
Select the genic value of probability corresponding gene position in the highest character string of described current fitness to tilt the gene of each gene position.
4, method as claimed in claim 3 is characterized in that, if the genic value of corresponding gene position is 1 in the highest character string of current fitness, then selects probability to increase the numerical value of appointment the gene of corresponding gene position; Genic value as if corresponding gene position in the highest character string of current fitness is 0, then selects probability to reduce the numerical value of appointment the gene of corresponding gene position;
Perhaps, if the genic value of corresponding gene position is 1 in the highest character string of current fitness, then select probability to reduce the numerical value of appointment the gene of corresponding gene position; Genic value as if corresponding gene position in the highest character string of current fitness is 0, then selects probability to increase the numerical value of appointment the gene of corresponding gene position.
5, the method for claim 1 is characterized in that, described termination condition perhaps is whether to reach default fitness threshold value for whether arriving default generation number threshold value.
6, a kind of quadrature mirror filter bank parameter optimization device based on genetic algorithm, it is characterized in that, comprise: random number generation unit, gene code unit, fitness computing unit, memory cell, gene are selected probability updating block and judgment processing unit, wherein
The random number generation unit is used for generating at random the random number that many groups are represented the quadrature mirror filter bank parameter;
The gene code unit is used for selecting the current gene of probability updating block to select probability according to gene, many groups random number of described random number generation unit generation is encoded to the string of binary characters of finite length respectively;
The fitness computing unit, the described string of binary characters of the described gene code unit that is used for decoding also calculates fitness, and the string of binary characters of high fitness that will calculate stores described memory cell into;
Memory cell has been used to store the highest string of binary characters of fitness since the successive dynasties;
Gene is selected the probability updating block, and the genotype that is used for the string of binary characters of the highest fitness that calculates according to described fitness computing unit is upgraded gene and selected probability, and sends decision instruction to described judgment processing unit;
The judgment processing unit is used for judging whether to satisfy termination condition according to described decision instruction, if, then decode having the string of binary characters of high fitness in the described memory cell, obtain the parameter of described quadrature mirror filter bank; Otherwise, start described random number generation unit and generate random number.
7, device as claimed in claim 6 is characterized in that, it is 0.5 that described gene selects the initial gene of each gene position in the probability updating block to select probability.
8, device as claimed in claim 6 is characterized in that, described gene selects the probability updating block to select the genic value of probability corresponding gene position in the highest character string of fitness that described fitness computing unit calculates to tilt the gene of each gene position.
9, device as claimed in claim 8 is characterized in that, if the genic value of corresponding gene position is 1 in the highest character string of current fitness, then described gene selects the probability updating block to select probability to increase the numerical value of appointment the gene of corresponding gene position; Genic value as if corresponding gene position in the highest character string of current fitness is 0, and then described gene selects the probability updating block to select probability to reduce the numerical value of appointment the gene of corresponding gene position;
Perhaps, if the genic value of corresponding gene position is 1 in the highest character string of current fitness, then described gene selects the probability updating block to select probability to reduce the numerical value of appointment the gene of corresponding gene position; Genic value as if corresponding gene position in the highest character string of current fitness is 0, and then described gene selects the probability updating block to select probability to increase the numerical value of appointment the gene of corresponding gene position.
10, device as claimed in claim 6 is characterized in that, also comprises:
Counting unit is used for the number of times of described random number generation unit generation random number is counted;
Described judgment processing unit, is judged and is satisfied termination condition during more than or equal to default frequency threshold value in the count value of described counting unit.
11, device as claimed in claim 6 is characterized in that, also comprises:
The fitness comparing unit, the fitness that is used for the highest string of binary characters of fitness that described fitness computing unit is calculated compares with default fitness threshold value;
Described judgment processing unit compares the highest described fitness when being equal to or greater than default fitness threshold value at described fitness comparing unit, judges and satisfies termination condition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200710119030 CN101068108A (en) | 2007-06-18 | 2007-06-18 | Orthogonal mirror image filter group realizing method and device based on genetic algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200710119030 CN101068108A (en) | 2007-06-18 | 2007-06-18 | Orthogonal mirror image filter group realizing method and device based on genetic algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101068108A true CN101068108A (en) | 2007-11-07 |
Family
ID=38880563
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200710119030 Pending CN101068108A (en) | 2007-06-18 | 2007-06-18 | Orthogonal mirror image filter group realizing method and device based on genetic algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101068108A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102664646A (en) * | 2011-05-17 | 2012-09-12 | 杭州畅鼎科技有限公司 | Filtering method for optimizing parameters by adopting genetic algorithm and nonlinear convex programming theory |
| CN103793747A (en) * | 2014-01-29 | 2014-05-14 | 中国人民解放军61660部队 | Sensitive information template construction method in network content safety management |
| CN109977227A (en) * | 2019-03-19 | 2019-07-05 | 中国科学院自动化研究所 | Text feature, system, device based on feature coding |
-
2007
- 2007-06-18 CN CN 200710119030 patent/CN101068108A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102664646A (en) * | 2011-05-17 | 2012-09-12 | 杭州畅鼎科技有限公司 | Filtering method for optimizing parameters by adopting genetic algorithm and nonlinear convex programming theory |
| CN102664646B (en) * | 2011-05-17 | 2014-07-02 | 杭州畅鼎科技有限公司 | Filtering method for optimizing parameters by adopting genetic algorithm and nonlinear convex programming theory |
| CN103793747A (en) * | 2014-01-29 | 2014-05-14 | 中国人民解放军61660部队 | Sensitive information template construction method in network content safety management |
| CN103793747B (en) * | 2014-01-29 | 2016-09-14 | 中国人民解放军61660部队 | A kind of sensitive information template construction method in network content security management |
| CN109977227A (en) * | 2019-03-19 | 2019-07-05 | 中国科学院自动化研究所 | Text feature, system, device based on feature coding |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| TWI767922B (en) | Method and apparatus for neural network quantization, and method of manufacturing chipset | |
| TWI801356B (en) | Method for network quantization, apparatus in neural network, method of manufacturing chipset, and method of testing apparatus | |
| US9058564B2 (en) | Controlling quarantining and biasing in cataclysms for optimization simulations | |
| CN1906855B (en) | Dimensional vector and variable resolution quantisation | |
| CN1783729A (en) | Method and apparatus for decoding low density parity check code using united node processing | |
| WO2011135410A1 (en) | Optimization technique using evolutionary algorithms | |
| CN101079639A (en) | Ldpc decoding apparatus and method based on node memory | |
| CN107609648B (en) | Genetic algorithm combined with stacking noise reduction sparse automatic encoder | |
| CN111079899A (en) | Neural network model compression method, system, device and medium | |
| CN101079642A (en) | Method for decoding error correcting code, its program and its device | |
| CN109918307A (en) | Path coverage test data creation method based on Negative Selection genetic algorithm | |
| CN101068108A (en) | Orthogonal mirror image filter group realizing method and device based on genetic algorithm | |
| CN114118369A (en) | Image classification convolution neural network design method based on group intelligent optimization | |
| Haseyama et al. | A filter coefficient quantization method with genetic algorithm, including simulated annealing | |
| CN114882945A (en) | Ensemble learning-based RNA-protein binding site prediction method | |
| US20100228808A1 (en) | Restrained Vector Quantisation | |
| KR20220004945A (en) | Scaffold-based molecular design with a graph generative model | |
| CN111652373B (en) | Task scheduling method, device and equipment of cloud robot and storage medium | |
| KR20200023695A (en) | Learning system to reduce computation volume | |
| US11488007B2 (en) | Building of custom convolution filter for a neural network using an automated evolutionary process | |
| CN1815940A (en) | Method for checking and correcting meta-viterbi algorithm error in communication channels and system thereof | |
| CN116978462A (en) | Method for generating non-natural promoter based on diffusion model | |
| CN116597894A (en) | XGBoost feature selection and deep learning combined gene-to-phenotype prediction method and system | |
| CN116305939A (en) | High-precision inversion method and system for carbon water flux of land ecological system and electronic equipment | |
| KR101369064B1 (en) | Audio encoding device and audio encoding method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |